[time-nuts] Temperature controlled TCVCXO
David
davidwhess at gmail.com
Sat May 14 00:46:57 UTC 2016
Thanks for this. It is exactly the kind of response I was looking
for. Based on what you say, it seems like a $10 eBay OCXO will always
beat an temperature controlled TCXO.
The varactor tuning range is going to be a problem for which the only
solution is a discrete crystal oscillator. I wonder if there are any
non-custom sources for suitable crystals in a design like this.
I assumed that they used AT cut crystals with a temperature curve of
the same form if not the same uncompensated temperature range. Of
course the temperature compensation circuit should completely dominate
that anyway.
I was hoping for something with a better risk for cost even if not as
good as the typical Ebay used OCXO of questionable heritage.
On Fri, 13 May 2016 17:57:58 -0400, you wrote:
>Ok, with a TCXO you have a temperature sensor that tries to servo the crystal on to frequency.
>You also have a crystal with a temperature dependance. As you try to heat / cool the TCXO
>your thermal variation hits one before it hits the other. The net effect is that the ADEV is actually
>worse with a TCXO than with an un-compensated crystal. Consider that a good oven has variation
>in temperature on the order of micro C over a few seconds to tens of seconds. Thats what you
>are trying to emulate.
>
>The next issue is that the TCXO has a loop design oriented towards wide tune range. Things like
>varicap diodes have a much bigger impact on a TCXO loop than they do on a narrow tune range
>OCXO loop. That (and possibly a coil or two) are yet another source of thermally induced variation
>in the oscillator. Just like the temperature sensor, there is a delay in the temperature hitting them
>compared to other parts of the oscillator. Again, you get a degradation in ADEV over the bare
>TCXO crystal.
>
>Next up is the crystal its self. A TCXO crystal likely has a turn temperature that is optimized to
>match the compensation approach being used. In some cases there will not actually be a turn
>at all. Thus there is no temperature zero slope point to hit with your temperature controller. This
>means that rather than you now need much better temperature variation control on your system.
>
>Bottom line: A $10 eBay OCXO is likely to beat an ovenized or cooled TCXO.
>
>Bob
More information about the Time-nuts_lists.febo.com
mailing list